The present invention provides a
pulse noise filtering method for load switch event detection which is used to solve the problem that a
common noise elimination device cannot effectively filter noisewith non-stationary and non-
Gaussian features in the prior art. The power
signal noise filtering method comprises the steps of: inputting actually measured power signals; performing segmentation on the actually measured power signals and rearranging data to a matrix; transforming the power
signal matrix; performing matrix initialization prior to iteration; and performing iteration calculation on the matrix after the initialization is completed to obtain a matrix after
pulse noise is filtered, and further converting the matrix to the power signals. Through transformation for the power
signal matrix, the sparsity of the power signals is processed to effectively filter the
pulse noise in the power signals and achieve
high fidelity for the useful power signals, the
phase step features of the power signals cannot be damaged, the method is simple and practicable, the computing speed is fast, and therefore, the accurate switch event detection can be performed to perform accurate decompositionfor domestic
electricity load.